How does embroidery work on clothes?

Machine embroidery stitches a design into a garment with needle and thread instead of printing on top of it. The artwork is first digitised, which means redrawing it as stitch paths, stitch types, directions and densities, because an embroidery machine reads coordinates and not pictures. The garment is then hooped tight over a stabiliser, and a multi-needle machine sews the design one colour at a time, moving the frame under a fixed needle. Embroidery is the most durable and most substantial decoration available on cloth, but it cannot do gradients, photographs or fine detail, and every design has a minimum size below which it stops reading.

What machine embroidery actually is

An industrial embroidery machine is a lockstitch machine with a moving frame. The needle takes the top thread down through the cloth, a rotary hook catches the loop underneath and locks it against the bobbin thread, the frame holding the garment shifts a fraction of a millimetre, and it happens again. A commercial head runs somewhere around 700 to 1,000 stitches a minute, slower on caps, metallic thread and dense satin work.

The needle does not move sideways. The garment does, and every part of the design is stored as points the frame travels between.

Nothing is printed. No ink, no curing, no chemistry. The decoration is thread held in the cloth mechanically, which is why it outlasts every printed finish, why it adds weight and texture rather than none, and why it has hard limits printing does not. Three of those limits run through this article: the machine understands only coordinates, so artwork has to be translated by a person; thread has real thickness and tension, so it pulls the fabric it is sewn into; and the garment has to be held flat and drum tight, which decides where a design can physically go.

What digitising is, and why it is not a file conversion

Digitising turns artwork into a stitch file. It is not exporting a logo to another format, and it is the part of the process customers most often underestimate.

A digitiser rebuilds the design object by object. For every shape they set the stitch type, the stitch direction, the density, the underlay beneath it, and how much extra width to add so the shape closes up correctly once sewn. They also set the sequence: what sews first, what sews over the top, and how to order the whole design so the machine makes as few colour changes and travel moves as possible. A left chest logo can take a couple of hours of that. A complicated back piece takes considerably longer.

Software can auto-digitise and the result is usually poor. Automatic tracing has no view on whether a 1.5 mm line will survive, no idea that the fabric is a loose knit that will swallow it, and no reason to sequence anything sensibly.

Formats matter more than people expect. A Tajima DST file, the common production format, stores only stitch coordinates and machine codes for jump, trim, colour change and end. It carries no thread colours and no shapes, so it cannot meaningfully be edited afterwards. Real editing happens in the native object file the digitiser worked in, which is why that file is worth keeping.

A stitch file is also tied to one size and one fabric. Scaling it up spreads the same number of stitches over a larger area, so the design goes thin and gappy. Scaling it down crams them into less space, so it goes hard and boardy. Beyond roughly ten percent either way, a design is redigitised rather than resized, and the same logo going onto a cap front, a brushed fleece and a light jersey properly wants three versions. The compensation is that digitising is done once and reused on every repeat order for the life of that logo.

The three stitch types, and what each is for

Almost every embroidered design is built from three stitches.

Satin stitch. Long stitches laid side by side across a shape, so the thread runs the width of the column and catches the light along its length. This is what gives embroidery its sheen, and it is used for lettering, outlines, borders and narrow shapes. It works from about 1 mm wide up to about 8 to 10 mm. Below 1 mm there is not enough thread to cover the cloth. Above 10 mm the stitches are long enough to snag on a fingernail and go loose in the wash, so wide columns are split into sections or converted to fill.

Fill stitch. Also called tatami. Rows of short stitches in a repeating pattern to cover a large area, with the row ends staggered so no line forms across the shape. This is how a crest, a solid block or a big letter is filled. Fill is flatter and more matt than satin, can carry a stitch angle and a texture, and is where stitch count accumulates fastest.

Running stitch. A single line of stitches used for outlines, fine detail and travel between areas. Doubled back on itself it becomes a bean or triple stitch, heavy enough to read as a drawn line.

The choice is not cosmetic. A 3 mm high letter drawn as satin fills in and turns to mush. It has to be enlarged, redrawn as running stitch, or dropped. That call is made before the machine runs, which is why a digitiser needs the final width of the design and the fabric it is going on.

Underlay, pull compensation and why designs distort

Two invisible parts of a stitch file do most of the work of making it look right.

Underlay is stitching sewn underneath the visible stitching. It stabilises the fabric before the top stitches land, stops them sinking into a pile or a loop, and gives the design its loft. A centre run underlay stiffens a narrow satin column, an edge run holds the borders, a lattice supports a fill. On brushed fleece and loopback terry, underlay is the difference between a crisp logo and one that looks like it is drowning.

Pull compensation exists because thread pulls. Every stitch carries tension, and a satin column draws the fabric inward along the stitch direction while pushing it outward at the ends. Sew a shape at its true drawn size and it comes out narrower than drawn across the column and slightly longer along it. A circle sews as a short ellipse, and an outline no longer sits on the fill it is meant to outline. The fix is to draw the shapes deliberately wrong: the digitiser widens the columns slightly, often around 0.2 mm each side, so the finished shape measures what the artwork says.

Knit is where this becomes obvious. Woven twill and canvas barely move. Jersey, rib and loose knits move a great deal, because the fabric has stretch of its own and the loop structure gives way under thread tension. On knit the same design needs more compensation, lighter density and a stronger stabiliser. Without that, three things go wrong at once: shapes come out squashed, outlines land off register, and the panel puckers into a ridge that never washes out. Puckering on knit is almost always a compensation and stabiliser failure rather than a machine fault, so fabric weight and structure are worth settling early using our streetwear fabric guide.

Stabiliser, backing and why hooping limits placement

Stabiliser, or backing, is the material behind the fabric that makes the fabric behave while it is sewn. It is chosen for the cloth, not the design.

• Cut-away backing stays in the garment for life, trimmed close after sewing, supporting the stitches every time the garment stretches or is washed. Knit needs cut-away, in a weight matched to the stitch count.
• Tear-away backing is removed after sewing and suits stable wovens, twill and cap fronts, where the fabric does not move on its own.
• Water-soluble topping sits on top rather than behind and washes out afterwards. It stops stitches sinking into terry, brushed fleece and knitwear, so the design sits on the surface rather than in it.
• Adhesive and fusible backings are for pieces that cannot be hooped directly and have to be floated on a hooped backing instead.

Hooping is the other half. The area to be embroidered is clamped between two rings, or into a magnetic or clamp frame, and pulled taut. That frame then travels under the needle, so the hoop area and its arms all have to sit flat.

This is the real constraint on placement. A hoop needs a flat, seam-free area big enough for the design plus a margin, and the garment has to open out enough to get onto the frame. A left chest, a centre chest, a back, an upper sleeve and a cap front are straightforward. A placement crossing a side seam, a zip, a pocket edge, a cuff or a hem is difficult or impossible on a finished garment, because the hoop cannot lie flat over the bulk. Cuffs, collars and small parts need small tubular or clamp frames, which a factory either has or does not.

So settle placement before the garment is designed rather than after. On a cut-and-sew run a panel can be embroidered flat before assembly, which removes the constraint altogether. That is a practical advantage of making garments from cloth instead of decorating blanks.

Needles, colour changes, trims and jump stitches

A commercial head carries several needles, typically 6, 9, 12 or 15, each threaded with a different colour. Multi-head machines run several heads in parallel from one file, so a run is sewn simultaneously with identical results.

Colour changes happen automatically as long as the design uses no more colours than the machine has needles. The head stops, the needle case slides across, sewing continues. There is no per-colour set-up in the sense screen printing has one, since there is no screen to make, but each change costs seconds of machine time. Where colour count starts to hurt is beyond the needle count, because then the machine has to stop and be re-threaded partway through, on every garment.

Jump stitches and trims are the hidden cost. Moving from one area to another, the machine either jumps, dragging a loose thread across the gap, or trims and ties off. Modern machines trim automatically, and a well-sequenced file keeps trims down by ordering shapes sensibly. A design made of many small scattered elements generates hundreds of trims, each a pause, each leaving a tail to be cleaned by hand. Finishing labour is real: trims clipped, backing cut away, topping washed out, piece pressed.

Thread types, and 3D puff on caps

Most production embroidery uses 40 weight thread, in one of two fibres. Polyester is strong, colourfast, resistant to chlorine and bleach, and survives industrial washing, so it is the default on anything worn hard or sent through a garment wash after decoration. Rayon is softer with a higher sheen and reads as more luxurious, but it is weaker and less bleach resistant, so it suits pieces treated gently.

Finer 60 weight thread pushes small lettering slightly below the usual limits at the cost of speed. Metallic threads look striking and run slowly and fragilely, needing lower speeds and a larger needle. The bobbin underneath is normally a fine white or black polyester, which is why the reverse of an embroidered piece is a different colour.

3D puff is a technique rather than a thread. A sheet of embroidery foam is laid on the garment, the design is sewn through it with dense satin columns, and the surplus foam is torn away with the last traces removed by heat. The result stands proud of the cloth, which is why it is the standard look on structured cap fronts. It has its own rules: columns have to be wide enough to hold the foam, roughly 4 mm and up, so fine detail and small text are out; density is raised; and the end of each column needs capping stitches to perforate the foam so it tears cleanly instead of showing at the edge. Designs are simplified and made bolder for puff, and it needs a firm base underneath, so it looks slack on a soft unstructured panel.

What embroidery is bad at

Embroidery is the most durable decoration here and the one that reads as most substantial. It is also bad at a specific and predictable set of things.

• Fine detail. Thread has physical width. Detail below roughly 1 mm cannot be represented, and below about 1.5 mm it is unreliable on anything but stable woven cloth. Hairlines, thin outlines and tiny gaps between shapes close up and merge.
• Small text. The practical floor for legible satin lettering is around 5 mm capital height for a simple sans-serif, with 4 mm the absolute limit on stable woven fabric. Serifs, script tails, thin strokes and condensed type all raise it. A strapline that sits happily under a printed logo usually has to grow or go.
• Photographic imagery. Thread is opaque and comes in discrete colours. There is no halftone and no dot, so a photograph can only be interpreted as a handful of flat colours, which is a different image.
• Gradients and blends. There is no way to fade one colour into another except by blending densities across a boundary, which is slow, specialist and only approximate at large sizes. Assume flat colour.
• Large solid areas. The one people underestimate. Filling a big area means tens of thousands of stitches, and all that thread has weight and stiffness. The panel goes board-like, the garment hangs badly, the area stops stretching with the cloth, and the piece feels heavy to wear. A full back in solid fill is nearly always wrong. Use an outline, a partial fill, appliqué over a fabric panel, or a printed base with an embroidered accent.

The practical rule: a design bound for embroidery usually has to be redrawn, not merely digitised. Redrawing means simplifying shapes so they hold at the final width, thickening strokes that fall under the minimum, flattening gradients to solid colours, cutting the colour count, enlarging or removing text, opening gaps that would close, and deciding what to lose. Judge the artwork at finished size, which for a left chest is usually around 60 to 90 mm wide. Print it at that size on paper and look at it before anyone digitises anything. If it does not read on paper it will not read in thread. Say what has to survive and what can go. A good digitiser will make those calls, but they should be your calls.

Stitch count, and what drives the cost

Stitch count is the number of stitches in the finished file, and it is the number that matters most, because it is the honest measure of how long a design occupies a machine head. A small left chest logo commonly lands somewhere around 5,000 to 12,000 stitches. A large back piece with solid fill can run into many tens of thousands.

Two things about it are worth internalising. It scales with area, not width: doubling the width of a design roughly quadruples the stitch count, so a logo that is quick at 70 mm is a different job at 140 mm. And it is driven by coverage rather than visual complexity, so an intricate outlined design can be lighter than a plain solid block. Designers routinely assume the opposite.

We do not publish numbers, but the cost structure is straightforward and the same everywhere. What moves an embroidery quote, roughly in order:

• Stitch count per placement, because it sets machine minutes per garment.
• Number of placements. Each is a separate hooping, handling and run. Two 6,000 stitch placements cost more than one 12,000 stitch placement, even though the totals match.
• Colour changes, as stops and machine time, and much more so once a design needs more colours than the head has needles and has to be re-threaded mid-run.
• Hooping difficulty. Cap fronts, sleeves, cuffs, collars and small components need specialist frames and slower handling. An awkward placement is a labour cost, not a stitching cost.
• Digitising, once per design per size and fabric, then reused for the life of the logo. It is a set-up cost, so it weighs on a first run and disappears into the background on repeats.
• Materials and finishing. Cut-away backing, water-soluble topping, foam for puff, speciality thread, and the hand trimming and pressing afterwards.

Quantity behaves as it does everywhere: set-up is fixed per design, stitching is per unit, so the per-piece figure falls as the run grows and then flattens, because machine minutes never go away. That is the same curve described in our explanation of minimum order quantities.

Embroidery against screen print, DTG and sublimation

The artwork and the fabric usually decide this together, before anyone thinks about method.

Read it as a shortlist rather than a verdict. A logo on fleece, twill or a cap goes to embroidery unless something rules it out. A photograph never goes to embroidery. A large all-over graphic on polyester goes to sublimation, which has no hand feel at all and is the opposite of embroidery on every axis. Flat spot colours on cotton in volume go to screen print. Mixing methods on one garment is normal and often right: an embroidered chest mark over a printed body, or a printed graphic with an embroidered detail.

Weight is the tiebreaker people forget. Embroidery flatters heavy cloth and fights light cloth. On a 400 GSM heavyweight hoodie it looks and feels right. On a light single jersey tee a heavy stitch count drags the fabric out of shape whatever the stabiliser, so the design has to be smaller and lighter, or printed instead.

Embroidery at NO FUGAZI

NO FUGAZI is a London-managed custom clothing manufacturer producing at its own dedicated facility. We are cut and sew, so garments are made from cloth rather than bought as blanks and decorated. For embroidery that matters in one specific way: a panel can be embroidered flat before the garment is assembled, which removes the hooping constraint on placements that would be impossible on a finished piece.

Digitising is done in house and the file is kept, so repeat orders sew from the same file and match the first run. A logo going onto more than one fabric or at more than one size is digitised properly for each rather than scaled and hoped over. Embroidery sits alongside 3D and puff embroidery, appliqué, chenille patches, reflective and glow-in-the-dark embroidery, reflective piping, screen print, high-density silicone, puff print, DTG, DTF, vinyl, sublimation, rhinestone and laser engraving. Most real jobs use more than one.

You do not need a tech pack to start. A reference image, a sketch, a photo of a garment you own, or a logo file and a note of where it should sit is enough, and the technical side gets drafted from that. The minimum order is 50 pieces per style. Sampling is not charged as a separate line when you commit to the bulk run with a deposit, with one revision round included, while a standalone sample without a bulk commitment is charged separately. Sampling matters on embroidery, because a sew-out on the actual fabric is the only honest way to approve a digitised file.

As ranges, sampling takes about 2 to 3 weeks and bulk about 3 to 8 weeks after sample approval, plus shipping. Quality control runs in four stages, fabric is pre-shrunk to 1 to 3 percent residual shrinkage, and we ship worldwide. Send the logo as vector art if you have it, the finished width, the placement, the fabric and the quantity. WhatsApp is fastest, on +44 7537 134878, or use the brief form. If you are still choosing a factory, our guide to finding a clothing manufacturer in the UK covers what to ask before sending artwork to anyone.

Questions

How does machine embroidery work?

An industrial embroidery machine holds a garment in a hoop and moves that hoop under a fixed needle. The needle takes the top thread through the cloth and a rotary hook locks it against a bobbin thread underneath, forming a lockstitch. A digitised file tells the machine every coordinate, stitch type, colour change and trim in sequence. Nothing is printed, so the decoration is thread held mechanically in the fabric.

What is digitising in embroidery?

Digitising is redrawing artwork as stitch instructions. A digitiser rebuilds each shape by hand, choosing stitch types, directions, densities, underlay, pull compensation and the sequence the machine sews in. It is not a file conversion, and automatic tracing produces poor results. The file is specific to one size and one fabric, and it is reused on every repeat order of that logo.

How small can embroidered text be?

Around 5 mm capital height for a simple sans-serif in satin stitch, and about 4 mm as an absolute floor on stable woven fabric. Below that the stitches close up and the letters fill in. Serifs, script, thin strokes and condensed type all need more height than that. On knit and fleece, add height again, because the surface swallows fine stitching.

Why does an embroidered logo pucker or look distorted?

Because thread pulls the fabric as it is sewn. Satin columns draw the cloth inward across the stitch direction and push it outward at the ends, so shapes come out narrower than drawn. A digitiser compensates by widening columns slightly, typically about 0.2 mm each side, and more on knit. Puckering is usually the wrong stabiliser, too high a density, or missing compensation on a stretchy fabric.

What determines the cost of embroidery?

Stitch count first, because it sets how long each garment occupies a machine head. Then the number of placements, since each one is a separate hooping and handling. Then colour changes, especially when a design needs more colours than the machine has needles and has to be re-threaded mid-run. Then hooping difficulty on caps, sleeves and small parts, plus a one-off digitising set-up per design.

What artwork do you need for embroidery?

Vector art is ideal, so AI, EPS or PDF, along with the exact finished width, the placement and the fabric. A high-resolution image works too, since the design gets redrawn as stitches either way. Name any colours that must match. Expect to simplify: fine detail, small text and gradients usually have to be enlarged, thickened, flattened or removed before it will sew well.

Related reading

• What is sublimation printing, and what is it good for? – the opposite method, with no hand feel and unlimited colour, and where it beats stitching
• Streetwear fabric guide – fabric weight and structure decide stabiliser, density and whether a design will pucker
• Minimum order quantity explained – how set-up costs and per-unit costs behave as a run grows
• What we make – the garments, fabrics and decoration methods available on a cut and sew run
• Screen printing vs DTG vs embroidery: how to choose – which decoration method suits your artwork, fabric and run length
• How to screen print t-shirts, at home and in production – the process in full, and where doing it yourself stops working
• What is sublimation printing, and what is it good for? – the process, why cotton fails, and how it compares with DTF